For developers frustrated with juggling multiple tools across their workflow, GitHub's latest Copilot upgrade might feel like getting an extra pair of hands. The company announced today that it's rolling out Agent Mode with Model Context Protocol (MCP) support to all Visual Studio Code users, expanding the AI assistant's capabilities beyond simple code completions to become a more autonomous coding partner.
Key Points
- Copilot Agent Mode now available in VS Code, supports MCP tools
- Pro+ plan offers 1,500 premium requests for $39/month
- Premium models include Claude 3.7 Sonnet, Gemini 2 Flash, GPT-4.5
- Code Review Agent and next-edit suggestions now generally available
The Agent Mode upgrade transforms Copilot from a passive assistant into an active collaborator that can autonomously identify relevant context, suggest terminal commands, and handle complex tasks across multiple files. What makes this particularly powerful is the integration with MCP, which GitHub CEO Thomas Dohmke describes as "a USB port for intelligence" that connects AI models to external tools and services.
In practical terms, MCP enables Copilot to access and utilize a developer's entire engineering stack – from database schemas to repositories – all through a standardized interface. For example, a developer could prompt Copilot to "Update my GitHub profile to include the title of the PR that was assigned to me yesterday," and Agent Mode would handle the entire workflow by identifying and calling the necessary tools iteratively until completing the task.
This release comes with substantial model upgrades as well. GitHub is making premium AI models from three major providers – Anthropic's Claude 3.5 and 3.7 Sonnet, Google's Gemini 2.0 Flash, and OpenAI's o3-mini – generally available through a new premium request system. These models complement unlimited access to the base model (currently OpenAI's GPT-4o) included in all paid plans.
To access these enhanced models, GitHub is introducing a tiered approach. Copilot Pro subscribers will receive 300 monthly premium requests beginning May 5, while Business and Enterprise customers will get 300 and 1,000 monthly requests respectively, starting between May 12-19. For power users, the company unveiled a new Pro+ plan at $39 per month that includes 1,500 premium requests and access to the most advanced models.
The company reports that Agent Mode is already showing impressive capabilities, achieving a 56.0% pass rate on SWE-bench Verified with Claude 3.7 Sonnet. GitHub expects these results to improve as reasoning models continue to advance. For developers who can't wait for the progressive rollout, manual activation is available now.
GitHub's announcement also includes the general availability of next edit suggestions for code completions, allowing developers to quickly implement suggested changes with a simple tab press, and the Copilot code review agent, which has already been used by over one million developers during its preview period.
The company is building an ecosystem around MCP with the release of an open-source GitHub MCP server that allows developers to add GitHub functionality to any LLM tool supporting the protocol. This standardization eliminates the need for custom integrations between AI models and tools, creating a more unified development experience.